The Theory of Critical Distances: A New Perspective in Fracture Mechanics

(NSIF)

CHAPTER 1

Introduction

Materials Under Stress

Publisher Summary

This chapter describes the state of the art in the prediction of material failure as articulated in national standards and specifications, and as used in practice in engineering companies. The chapter reviews the background material, symbols, and terminology related to materials under stress. A fundamental way to obtain information about the mechanical properties of a material is to record its stress–strain curve, usually by applying a gradually increasing tensile strain to a specimen of constant cross section. The chapter focuses on the deformation and failure of materials under stress, but emphasizes upon brittle fracture and fatigue including ductile fracture and certain tribological failure modes such as fretting fatigue. Material failure can be determined with precision only in two rather special cases. The first is simple tension, as described by the stress–strain curve: and the second is the propagation of pre-existing cracks, as described by linear elastic fracture mechanics (LEFM).. The use of computer-based methods such as finite element analysis (FEA) is also discussed. The chapter also highlights the use of traditional fracture mechanics and solid mechanics in failure prediction.

It is assumed that the reader is familiar with some basic theory regarding the mechanical properties of materials, as can be found in textbooks such as Ashby and Jones’ Engineering Materials (2005) or Hertzberg’s Deformation and Fracture Mechanics of Engineering Materials (1995), and also with the fundamentals of solid mechanics and fracture mechanics, for which many useful textbooks also exist (Broberg, 1999; Janssen et al., 2002; Knott, 1973). Nevertheless, in this chapter we will briefly review the background material and introduce symbols and terminology, which will be used in the rest of the book. We will be concerned, in general, with the deformation and failure of materials under stress, but emphasis will be placed on those types of failure which will be the main subjects of the book, especially brittle fracture and fatigue, but also including ductile fracture and certain tribological failure modes such as fretting fatigue. Of special interest from a mechanics point of view will be cracks, notches and other combinations of geometry and loading, which give rise to stress concentrations and stress gradients. In this respect, the use of computer-based methods such as finite element analysis (FEA) will also be discussed. We will finish with critical appraisal of the use of traditional fracture mechanics and solid mechanics in failure prediction, setting the scene for the developments to be described in the rest of this book.

1.1 Stress–Strain Curves

A fundamental way to obtain information about the mechanical properties of a material is to record its stress–strain curve, usually by applying a gradually increasing tensile strain to a specimen of constant cross section. Figure 1.2 shows, in schematic form, some typical results; note that here we are plotting the true stress (σ) and true strain (ε), thus taking account of changes in specimen cross section and length during the test. Most materials display a region of linear, elastic behaviour at low strains, and in some cases (line 1) this continues all the way to failure. This is the behaviour of classic brittle materials such as glass and certain engineering ceramics. More commonly, some deviation from linearity occurs before final failure (line 2). This non-linearity has three different sources: (i) non-linear elasticity, which is common in polymers; (ii) plasticity, that is the creation of permanent deformation, which occurs principally in metals and; (iii) damage, which is important in ceramics and composite materials. We will define the stress at failure in all cases as the maximum point in the curve, and refer to it as σu or the Ultimate Tensile Strength (UTS). In some cases (line 3) complete separation does not occur at σu, rather some reduced load-bearing capacity is maintained. This happens when damage such as splitting and cracking becomes widespread, for example in fibre composites. Finally, some stress–strain curves display other features (line 4) such as a drop in stress after yielding (in some metals and polymers) and a long post-yield plateau terminating in a rapid upturn in stress just before failure: this occurs in polymers which display plastic stability due to molecular rearrangements.

Fig. 1.2 Some typical stress–strain curves.

1.2 Failure Mechanisms

1.2.1 Failure at the atomic level

The study of failure mechanisms in materials has a tendency to get complicated, so it is worth remembering that, at the smallest scale, there are only two mechanisms by which materials can break, which I will call cleavage and tearing. Cleavage involves the fracture of atomic bonds; a crack can form by breaking the bonds linking two parallel planes of atoms, and this crack can then grow by the fracture of successive bonds near the crack tip, essentially unzipping the material in directions corresponding to atomic lattice planes. The fracture surface consists of a series of flat facets corresponding to the grains of the material. Tearing, on the other hand, occurs when material separates due to plastic deformation: atoms move around to create high levels of strain so that the material literally tears itself apart. This can manifest itself in various different ways, from macroscopic thinning (necking) or sliding (shearing) of material to microscopic void formation and growth. These two atomic failure mechanisms are often referred to as ‘brittle’ and ‘ductile’; however, I have avoided using these terms because they are also used with different meanings to describe failure modes at the macroscopic scale as discussed below.

1.2.2 Failure modes in engineering components

The failures of engineering components and structures occur by one of seven different modes: elastic, ductile, brittle, fatigue, stress-corrosion, creep, and wear.

Elastic failures are those failures which occur as a result of a low value of Young’s modulus, E. Two types of elastic failure can be mentioned. The first is excessive deflection, which may prevent the correct functioning of a structure – examples include bridges and vehicle suspensions. The second is buckling, by which, at a certain critical combination of load and elastic modulus, the deflections of a structure become unstable so that small deviations become magnified. A classic example is the collapse of a thin column loaded in compression.

Ductile fracture is the term used to describe failure occurring due to macroscopic plastic deformation; the material’s yield strength is exceeded over a large region so that plastic strain can occur throughout the load-bearing section, causing either fracture or a major change in shape so that the component can no longer function. In principle the prediction of this type of failure is simple, since the only consideration is that the stress in the part should exceed the yield strength. In practice, however, the spread of plasticity and the resulting redistribution of stresses and strains makes the prediction of plastic collapse loads a difficult analytical problem. For complex engineering structures, solutions are usually obtained using FEA and other computer simulations.

Brittle fracture refers to failures which occur as a result of rapid crack propagation. The crack in question may already exist (for example, in the form of a manufacturing defect or slowly growing fatigue crack) or it may form as a result of locally high stresses, for example near a notch. Once formed, the crack is able to grow, if the applied loads are high enough, by fracture of material near its tip. This material may fail by either cleavage or tearing. In classic brittle fracture, the process of crack growth is unstable, leading to almost instantaneous failure of the component. In such situations any plastic deformation is confined to the immediate vicinity of the crack, so there may be little sign of macroscopic plasticity. Figure 1.1 illustrates the difference between ductile and brittle failures in bolts tested in tension. This simple distinction between brittle and ductile fracture is complicated by the fact that intermediate situations can often arise: crack growth can occur more slowly and gradually, requiring a monotonically increasing load, if there is a significant amount of plasticity or damage near the crack tip. The study of crack propagation has created the science of Fracture Mechanics, which will be discussed in more detail below.

Fig. 1.1 Examples of ductile fracture (left) and brittle fracture (right) in bolts ( Wulpi, 1985).

Fatigue is a process of crack initiation and growth, which occurs as a result of cyclic loading. A regular cycle of stress, such as a sine wave (Fig. 1.3), can be described using two parameters: the stress range Δσ and the mean stress σmean. Another common descriptor is the load ratio R, defined as the ratio of the minimum and maximum stresses in the cycle:

Fig. 1.3 Definition of parameters for cyclic loading.

(1.1)

The most common type of fatigue test involves applying a cyclic stress to a test specimen and counting the number of cycles to failure Nf. Separation will occur when a crack has grown to a sufficient length to cause a ductile or brittle fracture of the remaining cross section: some workers prefer to define failure as the creation of a crack of a specified size, usually a few millimetres. Figure 1.4 shows typical stress-life curves, describing the dependence of Nf on Δσ and σmean. In some materials the curve becomes effectively horizontal for Nf values in the range 10⁶–10⁷ cycles, allowing one to define a fatigue limit, Δσo; often, however, there is no clear asymptote in which case the fatigue limit is defined at a specified number of cycles, when it is often called the fatigue strength. Recent work, which will be discussed further in Chapter 9, has shown that this asymptote can be somewhat illusory: in some materials failures can occur at very large numbers of cycles, in excess of 10⁹, at low values of stress range. Changing the mean stress or R ratio will shift the entire curve. If the applied stress is high enough to cause large-scale plastic deformation on every cycle, then a non-linear stress–strain relationship will occur, the nature of which may change during cycling as the material hardens or softens. In such cases it is common practice to use the strain range Δε as the characterising parameter, instead of Δσ. In this situation the number of cycles to failure is generally low: this type of fatigue is referred to as low-cycle fatigue (LCF) to distinguish it from high-cycle fatigue (HCF) which occurs under nominally elastic conditions.

Fig. 1.4 Typical fatigue stress-life curves.

In real engineering components a crack may already exist, in which case one is interested in how fast it is propagating. The crack growth rate is usually expressed in terms of number of cycles, da/dN, rather than time, because normally the amount of crack growth per cycle is rather insensitive to the cycling frequency; there are, however, important exceptions to this rule, especially among polymers. The crack growth rate has been found to be a function of ΔK, the range of stress intensity, K, which will be defined below in Section 1.4.

Figure 1.5 shows the typical dependence of crack growth rate on stress intensity range, which displays two asymptotes: a growth threshold, ΔKth, below which crack growth effectively ceases, and an upper limit where the conditions for rapid, brittle fracture are approached. Changing the mean K, or R ratio, shifts the curve as shown.

Fig. 1.5 Typical fatigue crack growth rate curves.

Stress-corrosion cracking (SCC) is a form of gradual failure which is rather like fatigue in that it proceeds by crack initiation and gradual propagation. However, in this case the crucial factor is not a cyclic stress but the existence of a corrosive chemical environment. The mechanisms of SCC are many and varied but usually involve some kind of synergistic action between the chemical process and the applied stress. This type of failure will not be discussed in any detail in this book; it is quite likely that the Theory of Critical Distances (TCD) could be used to predict failures that occur by SCC, but to date this has not been investigated.

Creep is a process of plastic deformation that occurs gradually. In fact all plastic deformation processes are thermally activated, proceeding more easily as the temperature is increased towards the material’s melting point. Creep failures can also involve the creation and growth of crack-like damage. Critical distance methods have been used to study creep (see Section 9.7) but not in enough detail to merit discussion in this book.

Wear is the general name given to tribological failures, that is to failures which occur due to the rubbing action between two surfaces. If compressive stress and a sliding (shear motion) occur across a material interface, then very high local stresses will arise due to small surface irregularities, creating conditions in which material can be removed from one or both surfaces. There are various mechanisms of wear: the one that will be of most interest to us is known as contact fatigue and involves the creation of cracks at or near the point of contact. These cracks can grow to cause removal of surface material by spalling, for example in gear teeth. If there are also cyclic body forces in the component, then cracks which are initiated by contact fatigue can subsequently grow into the component by conventional fatigue processes: this type of failure is known as fretting fatigue. The prediction of tribologically induced failures such as these is difficult owing to the problems involved in estimating local stresses, which are affected by surface roughness, surface deformation and lubrication.

1.3 Stress Concentrations

It is almost inevitable that, in any engineering component, stresses will vary from place to place, and that failure will occur in locations where stresses are relatively high. One can think of a few exceptions to this rule – wires and tie-bars under pure tensile loading, for example – but apart from these we can say that the phenomenon of stress concentration is responsible for all mechanical failures in practice. Stress concentration has two causes: loading and geometry. Loading modes which cause stress gradients include bending and torsion, both of which tend to concentrate stresses at the surface. But, as we shall see, this type of stress concentration is generally very mild in comparison to the effect of geometric features such as holes, corners, bends and grooves.

To illustrate the magnitude of stress concentrations, it is useful to consider a specific example. Consider a rectangular bar of material of width 30 mm – the length and thickness of the bar are not important. If loaded in tension with a stress of 100 MPa the stress at any point in the bar will, of course, be the same. If we introduce a central hole, of radius 3 mm, then the stress will become much higher at two points (the ‘hot spots’) on the circumference of the hole. In fact the local stress will be approximately 300 MPa, because the stress concentration factor Kt for a hole is 3 (actually this is the Kt factor for a hole in a body of infinite width, but it is close enough for our present purposes). Figure 1.6 shows how the stress decreases with distance from the hole, along a line drawn perpendicular to the hole surface (and therefore also perpendicular to the axis of principal stress). Also shown is the stress distribution that would arise if we were to replace the hole with a crack of the same size (i.e. one whose half-length a is equal to the radius of the hole). It is important to note that the stress being plotted here is the elastic stress, that is we have calculated the stress assuming that there will be no yielding or other non-linear deformation behaviour in the material. This is an assumption that we will return to later on. Finally, the graph also shows the stress distribution that will occur if, instead of introducing a hole or crack, we subject the bar to pure bending with the same nominal stress; in this case the graph shows stress as a function of distance from the surface of the bar at which the maximum tensile stress occurs.

Fig. 1.6 Examples of stress–distance curves showing the effect of geometric features (a crack and a circular hole of the same size) and of bending loads. In all cases the nominal stress is 100 MPa; stresses are plotted as a function of distance from the point of maximum stress.

It is clear that the stress concentration effect due to bending is much smaller than that caused by either of the geometric features. Of course, the gradient of stress in bending will increase if we decrease the width of the bar, but we would have to make the bar very narrow indeed to create the stress gradients caused by the hole and the crack. The maximum stress due to the crack is theoretically infinite because, having zero radius at its tip, it creates a singularity in the stress field. It is interesting to note that, whilst the crack causes much higher stresses in its immediate vicinity, the stresses due to the hole are actually higher than those for the crack at larger distances. We shall see later that this observation turns out to be very significant. In order to understand the effect of geometric features on mechanical failure, it is necessary to consider not only the maximum stresses which they create, but also how these stresses change with distance. Indeed the majority of the analyses conducted in this book will make use of these elastic stress–distance curves.

1.4 Elastic Stress Fields for Notches and Cracks

The study of stress concentration effects is mostly carried out using notches. As Fig. 1.7 shows, a notch can be defined by three parameters: its depth D, root radius ρ and opening angle α. To be precise one should add a fourth feature, the notch shape, to include the fact that the sides of the notch can have different amounts of curvature. However, in practice the two features which mostly control stress concentration are D and ρ, with notch angle having a secondary effect which becomes significant at large values (α > 90°). In considering notch stress fields we will normally use coordinates centred on the point of maximum stress, at the notch root: Fig. 1.7 shows a polar coordinate system (r, θ).

Fig. 1.7 Definition of parameters for notch geometry: length D, root radius ρ and angle α. Stresses are normally defined using cylindrical coordinates (r, θ) centred on the point of maximum stress at the notch root.

The reason that researchers use notches to study stress concentration effects is because they are relatively simple to make and test experimentally and to analyse theoretically. Much of the work described in this book (especially in Chapters 5–9 and 11) will be concerned with the effect of notches. However, it should not be forgotten that our real purpose in doing all this is to predict the behaviour of stress concentration features in engineering structures and components, which can be geometrically much more complex. For this reason we will consider some components in Chapter 12, along with features such as corners and joints, and in Chapter 10 we will consider stress concentrations which arise due to contact, causing fretting fatigue and other tribological failures. Defects such as porosity and inclusions also fall into the general category of stress concentrations.

To return to notches, some simple analytical solutions exist in certain cases. For example, the stress field created by a circular hole in a body of infinite size can be described as a function of applied nominal stress σ and hole radius ah. For the case of θ = 0 the result